From predictions to recommendations: Tackling bottlenecks and overstaying in the Emergency Room through a sequence of Random Forests

One of the goals to improve the quality of care in hospitals is to set a maximum of four hours for patients to be diagnosed and/or receive acute care in the Emergency Room (ER). Unfortunately, this is not always true and some patients overstay. The aim of this work is threefold: (1) to identify which patients will overstay during their admission to the ER; (2) to identify which (pair of) activities might heavily influence the time spent in the ER; and (3) to recommend actions to reduce such time. For that, a sequence of insightful supervised prediction models for generating recommendations is proposed. The method provided makes it possible to generate useful/actionable recommendations for problematic patients based on activities. State of the art techniques did not manage to generate recommendations at the arrival of the patient and/or did not take the interplay between patients into account.